LIN 1300 Chapter Notes - Chapter 16: Target Text, Speech Synthesis, Machine Translation
19 Oct 2016
School
Department
Course
Professor
Document Summary
Speech synthesis: using machine to produce human-like speech. Synthesized speech: piecing smaller record speech units into new utterances. Isolated word input: limit user to single word answer; pause after each word. Speaker dependent: user train system to recognize their voice only. Human translators are expensive, globalization causes rise in need of translation and not enough translators. Accurate translations at high speed and low cost. Translation: converting text in source language into text in target language. Not enough known about language and translation process to duplicate. Direct translation: every word translated, reordering based on morphological/syntactic rules of target language. Categorical divergence: translate into words that have different parts of speech. Conflational divergence: translation of +2 words into 1 word. Structural divergence: realization of verb arguments in different syntactic configurations. Interlingua translation: source language text translated into intermediate abstract representation containing sufficient info to allow creation of target language text. Creating common intermediate representation is hard, must know more about language first.
